CN1526236A

CN1526236A - Method and system for improving compressed image chroma information

Info

Publication number: CN1526236A
Application number: CNA028137922A
Authority: CN
Inventors: G��A��Ī˹; G·A·迪莫斯
Original assignee: Dolby Laboratories Licensing Corp
Current assignee: Dolby Laboratories Licensing Corp
Priority date: 2001-07-12
Filing date: 2002-07-12
Publication date: 2004-09-01
Anticipated expiration: 2022-07-12
Also published as: AU2002316676B2; SG10201508768WA; US20150358627A1; EP2493183A1; US7961784B2; US20120087408A1; EP2493179A1; ES2386872T3; US10504205B2; US9497471B2; US20120087409A1; US20080273809A1; JP2013093881A; SG143069A1; US9584816B2; EP2493181A1; EP2493184A1; SG193680A1; EP2493178A1; JP2013093885A

Abstract

Methods, systems, and computer programs for improving compressed image chroma information. In one aspect of the invention, a resolution for a red color component of a color video image is used that is higher than the resolution for a blue color component of the color video image. Another aspect includes utilizing a lower or higher value of a quantization parameter (QP) for one or more chroma channels as compared to the luminance channel. Another aspect is use of a logarithmic representation of a video image to benefit image coding. Another aspect uses more than two chroma channels to represent a video image.

Description

Be used to improve the method and system of compressed image chrominance information

Technical field

The present invention relates to image compression, especially relate in the image compression system of similar MPEG method, system and computer program in order to the chrominance information of improving compressed image.

Background technology

The MPEG background

MPEG-2 and MPEG-4 are the international video compression standards of regulation video sentence structure, and it provides a kind of effective means, with the form represent images order of compacter coded data.With the language of bits of encoded is " sentence structure ".For example, some signs can be represented a whole sampling block (as 64 samplings of MPEG-2).These two mpeg standards have also been described a kind of decoding (reconstruction) and have been handled, and wherein Bian Ma bit is converted to a kind of being similar to of unprocessed form of this image sequence by the expression from compactness.For example, mark in coded bit stream indicates that whether follow-up bit, will be earlier with a kind of prediction algorithm processing before being decoded with discrete cosine transform (DCT) algorithm.The algorithm that comprises decoding processing is by being adjusted by the semanteme of these mpeg standard definition.Can use this sentence structure and utilize common video properties, as spatial redundancy, time redundancy, uniform motion, spatial masking or the like.In fact, these mpeg standards have defined a kind of program design language a kind of data format of making peace.Mpeg decoder input traffic of must resolving and decode, but as long as data stream conforms corresponding M PEG sentence structure, can use possible data structure of all kinds and compress technique (departing from standard because of semantic discrepancy) though this sees technically.It also is admissible carrying required semanteme in a replacement sentence structure.

These mpeg standards use various compression methods, comprise in the frame and inter-frame approach.In most of video pictures, background keeps relative stability, and occurs action simultaneously in prospect.Background can move, but a large amount of picture is redundant.These mpeg standards are called as " inner (intra) " frame or " I frame " by generating one reference frame begins compression.The I frame is compressed with other frames irrelevant, therefore comprises a complete frame video information.The I frame provides the inlet point that enters a data bit stream in order to random-access, but it can only be by moderate compression.Typically, represent the data of I frame just to be placed on (though between the I frame, using much wide that interval also is effective in some cases) in the bit stream every 12 to 15 frames.Thereafter, owing to only have the sub-fraction frame to drop between the benchmark I frame different, so only there is image difference to be hunted down, to compress and store with bracket (bracketing) I frame.There is two types frame to be used to such image difference: the frame of prediction or P frame, and the frame of two-way interpolation or B frame.

The P frame is encoded with reference to the frame (an I frame or a previous P frame) in a past usually, and is often used as a benchmark for follow-up P frame use.The P frame is accepted quite high decrement.The B frame provides the highest decrement, but both needs the reference frame in a past also to need the reference frame in a future in order to be encoded.In standard compression techniques, two-way frame never is used to reference frame.

Macroblock is the zone of image pixel.For MPEG-2, a macroblock is 16 * 16 pixel groups with four 8 * 8DCT modules, and it has a motion vector that is used for the P frame, and one or two motion vectors that are used for the B frame.Macroblock in the P frame can use in the frame or interframe (prediction) is encoded and encoded one by one.Macroblock in the B frame can use intraframe coding, forward predictive coded, back forecast coding or forward direction and back to be encoded one by one to the predictive coding of the two (being two-way interpolation).But in the MPEG-4 video coding, use a kind of slightly different similarly structure.

After coding, a mpeg data bit stream comprises the sequence of I, P and B frame.A sequence can almost be made up of any pattern of I, P and B frame (having some small semanteme restrictions for its position).But, in industrial practice, have a fixing pattern (for example IBBPBBPBBPBBPBB) usually.

The MPEG color space is represented

MPEG-1, MPEG-2 and MPEG-4 utilize Y, U, the V color space to be used for compression.The formula that exists brightness to select can be expressed as but be converted to the typical conversion that YUV represents between RGB (R-G-B):

Y＝0.59G+0.29R+0.12B

U＝R-Y

V＝B-Y

Depend on color system, between 0.55 and 0.75, change corresponding to the Y luminance factor of green.The factor corresponding to redness changes between 0.2 and 0.3, and changes between 0.05 and 0.15 corresponding to the factor of blueness.

This conversion can be used as matrix conversion and is calculated, and this matrix conversion is to use the linear operator of design for linear signal.But this simple conversion is carried out with MPEG1,2 and 4 in the nonlinear video space, has just drawn different artefact (artifact) and problem.

The resolution that is typically reduction U and V chrominance channel in MPEG is to obtain bigger compression.It is that level is used 1/2nd resolution with vertical both that the most normally used resolution reduces.MPEG-2 supports full resolution colourity and level 1/2nd resolution.But, the most normally used MPEG-2 category (Profile), i.e. fundamental category (main profile at main level) on main stor(e)y time and the fundamental category (MP@HL) on high-level, level with vertical on all use 1/2nd resolution.The version 1 of MPEG-4 with 2 level with vertical on all use 1/2nd resolution.It should be noted that panchromatic degree resolution was commonly called 4: 4: 4,1/2nd colourity horizontal resolutions were commonly called 4: 2: 2, and 1/2nd vertical and horizontal resolutions were commonly called 4: 2: 0.(should be noted that term 4: x: x is leaky with regard to its implication and inference, but way being to describe the relation of chroma resolution to brightness with it usually).

When being converted to U and V color showing, be applied in the descend filter of low-level and vertical chrominance resolution of different mpeg standard for non-linear video signal.When using inverse conversion with recovery RGB, nonlinear signal and filter are to produce the mode reciprocation of artefact and problem.These problems can be summarized as " crosstalking " between Y brightness and U and V chrominance channel together with the space aliasing.

About the further information of linearity and nonlinear expression and conversion is found in " being used for the use (The Use of Logarithmic) of the logarithm and the density unit of pixel " that Gary Demos delivers in the SMPTE in October nineteen ninety meeting, this paper is published (in October, 1991 on the SMPTE periodical, volume 100, no.10)." upgradeable, the interoperable and extendible be used for the example (An Example Representation for Image Color andDynamic Range which is Scalable; Interoperable; and Extensible) that color of image and dynamic range are represented " that sees also that Gary Demos delivers in the SMPTE meeting in October, 1993, this paper is published on scientific research report collection and preprint.These file descriptions on the different phase of image Compression streamline the advantage of logarithm and linear space, and be combined in herein as the reference data.

Colourity sub sampling (sub-sampling)

Reason for U and V reduction chroma resolution is that the human visual system is more insensitive for the variation on the variation specific luminance Y on U and the V.Because Y mainly is green, U and V mainly are respectively red and blue, and this also can be said to be is that human vision sensitivity is red higher with blueness to the green comparison.But though U and V are considered to be identically in MPEG-1, MPEG-2 and MPEG-4, it is more responsive that the human visual system is compared to V (for its blue composition) for U (for its red composition).

Be embodied in the 1951NTSC-2 color standard that is used for television receiver in this difference aspect the colourity susceptibility.NTSC-2 uses the YIQ color space, and wherein I and Q are similar to U and V (having slightly different weighting).That is to say that the I passage represents that basically redness deducts brightness, and the Q passage represents that basically blueness deducts brightness.In NTSC-2, brightness is given the analog bandwidth of 4.5MHz, and the I chrominance channel is given the analog bandwidth of 1.5MHz.The Q passage of expression blueness-yellow axle then only is given the analog bandwidth of 0.5MHz.Therefore, it is its three times of giving the Q channel allocation that the NTSC-2 television system is given the information of I channel allocation, is to the I passage three times and give the information of Y luminance channel.Therefore, the bandwidth ratio between Y luminance channel and Q (blueness deducts brightness) passage is 9.These MPEGYUV and NTSC-2 correlation are summarized in the following table:

Chroma resolution is to the ratio of brightness

Ratio YUV YUV YUV NTSC-2

4∶4∶4 4∶2∶2 4∶2∶0

Red, U and I pixel were to Y 1: 12: 14: 13: 1

Blue, V and Q pixel were to Y 1: 12: 14: 19: 1

Obviously, in processing, the difference that exists between luminance channel and U and the V passage under mpeg standard is bigger than the difference of brightness under the NTSC-2 standard and I and Q interchannel.

Summary of the invention

The present invention proposes method, system and the computer program of the chrominance information that is used to improve compressed image.

Especially, in one aspect of the invention, color video frequency image can be used for red resolution (perhaps being used for the U resolution that YUV represents) that RGB represents by order and improve and surpass the resolution that is used for blueness (perhaps V) and improve.Use low resolution for example to mean for blue composition and in a motion compensation color video frequency image compressibility, need to compress less information.Of the present invention this comprises method, system and computer program on the one hand, the resolution of its red composition by selecting to make color video frequency image is higher than the resolution of the blue composition of this color video frequency image, and in video image compression system the image chroma information of compressed color video image.

Another aspect of the present invention is a kind of technology that is used to reduce the chrominance noise level and improves picture quality thus, and wherein chrominance noise is caused by arbitrary set-point of the quantization parameter that uses between compression period (Qp).This by for U (=R-Y) the passage utilization realizes than QP value lower outside the Y passage.Similarly, V (=B-Y) quality of passage can also be by improving than QP value lower outside the Y passage for the utilization of V passage.

Another aspect of the present invention be a kind of when the needs higher compression otherwise effective technique.At this on the one hand, the QP value of Y passage is applied a positive QP setover, for any one or two uses in U and the V chrominance channel.

Another aspect of the present invention is to use logarithm to represent to be beneficial to image encoding.If feasible, can improve the code efficiency that the yuv color space of the image that is expressed as linear rgb pixel value is at first represented to number encoder.On other treatment step, may be useful to the reciprocal conversion of linear expression.

Another aspect of the present invention be a kind of in image compression system the method in order to the video features that improves color video frequency image, comprise: select one group of image passage with the expression color video frequency image, comprise a luminance channel and n chrominance channel, wherein n is at least 3; And this luminance channel and n additional chrominance channel are compressed into a compressed video image.

In accompanying drawing and following specification, will set forth one or more embodiments of the detail of the present invention.To obviously see other characteristics of the present invention, purpose and advantage by specification and accompanying drawing and by claim.

Description of drawings

Fig. 1 is a flow chart, and it shows in yuv color space is represented in order to will the resolution of U be improved the illustrative method that surpasses the employed resolution of V (its can by computer realization).

Fig. 2 is a flow chart, and it shows the illustrative method (its can by computer realization) that applies QP biasing for chrominance channel.

Fig. 3 is a flow chart, and it shows for brightness and chrominance information and carries out illustrative method (its can by computer realization) to number encoder.

Fig. 4 is a flow chart, and it shows the illustrative method (its can by computer realization) of the additional chrominance channel that is used to encode in image compression system.Identical reference marker is represented identical unit in different accompanying drawings.

Embodiment

The color coding precision of improving

Along with the quality of image is improved with respect to characteristic such as the noise, the dynamic range of expansion and the Color Range of expansion that reduce, the people also improves the susceptibility of color.Especially have been noted that redness (the perhaps U in YUV represents) in RGB represents is usually than normally used needs in video compression more high accuracy and transparency.

Except that blueness is being used to handle (mixing or graphical analysis as the blue screen special effects), the mankind are for the susceptibility of the chrominance axis of blueness-yellow, as what embody, obtained suitable solution by the sampling of the half-resolution on level and the vertical direction by blueness or V.Four of the sum of all pixels of an image therefore ,/once providing enough qualities to be used to represent blueness or V chrominance axis.But with blue different with V, the coding of the half-resolution of redness and/or U is being not enough aspect the quality for big broad dynamic range display and projecting apparatus sometimes.

Therefore, image can be used for red resolution that RGB represents (perhaps being used for the U resolution that YUV represents) by order increases and surpasses blue (perhaps V) resolution and improve.Use lower resolution to mean the information that for example in a motion compensation color video frequency image compressibility, needs to compress still less for blue composition.

According to the present invention, have three kinds of blue (perhaps V) resolution of (downfiltered) with respect to descending filtration to keep the method for optimizing of redness (perhaps U) resolution of increase:

1) uses full resolution for redness and/or U;

2) only on a chrominance axis of vertical or horizontal direction, use 1/2nd resolution for redness and/or U; Perhaps

3) on one or two chrominance axises, use the resolution of filtering for redness and/or U, this resolution is between full size and 1/2nd, as 2/3 or 3/4.

Fig. 1 is the flow chart that shows a kind of illustrative method (its can by computer realization), and it is higher that this method represents that at yuv color space the resolution of using for U in (a kind of similar method can be applied to a RGB color space and represent) is compared to the resolution that V uses:

Step 101: utilize in the image compression system that yuv color space represents at one, with the V of an input picture (=B-Y) passage dwindles and is filtered into 1/2nd resolution on the horizontal direction, and optionally is 1/2nd resolution on the vertical direction.

Step 102: with the U of image (=R-Y) passage dwindle be filtered into one be higher than V (=B-Y) resolution of passage, preferably Xia Mian one of them:

A) full resolution;

B) horizontal direction between 1/2nd and full resolution between, but vertical direction is a full resolution;

C) level and vertically all 1/2nd and full resolution between;

D) vertical direction between 1/2nd and full resolution between, but horizontal direction is a full resolution;

Step 103: the compressibility compression YUV image (U and V chrominance information that it has brightness Y and dwindles filtration) that uses a similar MPEG.

Step 104: shorten image decompression into Y, U and V passage (normally in different computers).

Step 105: use suitable resolution increment (promptly in above-mentioned steps 101 in V and the above-mentioned steps 102 for the employed inverse that dwindles filterable agent of U), U and V passage are converted to full resolution.

Step 106: optionally be the RGB image that is used to watch, analyze or further handles with the YUV image transitions.

The different QP biasings that is used for colourity

Be called " the high accuracy Code And Decode of video image " and transfer the present invention (this with it in conjunction with as with reference to data) assignee's co-pending U.S. Patent application No.09/798,346, instructed the different aspect that between compression period, utilizes quantization parameter (QP).Thereby a further aspect of the invention is a kind of technology that the chrominance noise level is improved picture quality that is used to reduce, and this chrominance noise is that arbitrary set-point of the quantization parameter (QP) that uses between compression period causes.This by to U (=R-Y) the passage utilization QP value lower than Y passage realizes.Similarly, (=B-Y) quality can also be by improving for the V passage utilization QP value lower than Y passage for V.

A kind of straightforward procedure that realizes reducing colourity QP value is to deduct a steady state value from the QP value that is used to Y (brightness) passage.Alternatively, an independent steady state value (lower than the QP value that is used for Y) can be used for each U and V.For example, can deduct " 2 " from the QP value that is used for Y, thereby draw the QP value that is used for U, and can deduct " 1 ", thereby draw the QP value that is used for V from the QP value that is used for Y.Can adopt the QP value of being used to deduct any effective numerical value, this numerical value only is limited as minimum value " 1 ".

This method is effective for constant QP value (variable bit rate).It is effective too for variable QP value (for example in the motion compensation compressibility of constant and variable bit rate), this is because the QP value of moment can be setovered by the difference that deducts a regulation from the QP value that is used for Y, thereby draws the QP value that is used for each U and V.

In addition, as what describe in the patent application " the high accuracy Code And Decode of video image " mentioned in the above, the scope of the QP value of these different colourities biasings can be utilized the QP codomain function of expansion or search and expanded.

Except that using prearranged numerical value, must from the encoder to the decoder, signal the bias of U and V.They for example can be talked with corresponding to each, image sets (GOP), frame or image-region regulation once.

Fig. 2 is one the flow chart that applies the illustrative method (its can by computer realization) of QP biasing for chrominance channel is shown:

Step 201: in an image compression system, be used for the QP value of each U and V chrominance channel with a set point value (it can be different for each passage) reduction.

Step 202: the QP value with this reduction is used for U and the compression of V chrominance channel respectively.

Step 203: optionally,, guarantee that then the U and the VQP value that reduce are " 1 " at least if used variable QP value.

Step 204: except that using the biasing of presetting, (it once was minimum value) signaled or transmitted the reduction of QP value and measure to decoder when the QP value may change.

Step 205: use above-mentioned suitable QP value decompression (normally in different computers) signal for U and V (guaranteeing that once more the QP value that reduces is " 1 " at least).

Step 206: optionally, processing or the analysis of watching the image of decompression maybe this image to be used to add.

Another aspect of the present invention be a kind of when the needs higher compression otherwise effective technique.At this on the one hand, a positive QP biasing is applied to the QP value of Y passage, for use in any one or two in U and the V chrominance passband.(preferably checking if any, the QP maximum of compressibility).Can use independent bias to each U and V passage.In addition, the step of such an embodiment is similar to those steps shown in figure 2.

Brightness and colourity to number encoder

The paper of mentioning in the above that is called " for the logarithm of pixel and the use of density unit " has been described the advantage that logarithm is represented for dynamic range.Represent how much to be similar to normally used video translation function with the logarithm of dynamic range coupling.Although similar, to compare with normally used various representation of video shot, it is all more suitable aspect the orthogonality of chrominance channel that logarithmic table is shown in ductility, calibration is used, reached.

Another aspect of the present invention is to use logarithm to represent to be beneficial to image encoding.If have been found that the code efficiency that the feasible yuv color space that can improve the image that is expressed as linear rgb pixel value (as on the transducer of camera) at first of number encoder is represented.On other treatment steps, the reciprocal conversion of linear expression may be useful.

As what describe in the patent application " the high accuracy Code And Decode of video image " mentioned in the above, the colourity of brightness crosstalked can reduce to minimum at this moment:

Ylog＝Log(Wr*R+Wg*G+Wb*B)

U＝Log(R)-Ylog

V＝Log(B)-Ylog

Wherein Wr, Wg and Wb are the linear weighted functions of the redness, green and the blue composition that are used for brightness, and wherein R, G and B represent a linear light space (light space).These relational expressions are useful using of the present invention this on the one hand.

Fig. 3 is the flow chart that shows an illustrative method (its can by computer realization), this method system be used for brightness and chrominance information to number encoder:

Step 301: in an image compression system, carry out following conversion for (for example directly from a video camera) linear R, the G and the B pixel value of input:

Ylog＝Log(Wr*R+Wg*G+Wb*B)

U＝Log(R)-Ylog

V＝Log(B)-Ylog

Wherein Wr, Wg and Wb are the linear weighted functions of the redness, green and the blue composition that are used for brightness.

Step 302: the resolution (as mentioned above) that optionally reduces U and V chrominance channel.

Step 303: this Y, U and V for live image represent to carry out the motion compensation compression.

Step 304: with the image decompressor compressed Y, U and V color composition with the image that reactivates (normally in different computers).

Step 305:, the reduction of above-mentioned resolution is fallen back to recover U and V resolution completely to returning if used selectable step 302.

Step 306: use following conversion to recover linear R, G and B pixel value:

The contrary log (Y+U) of R=

The contrary log (Y+V) of B=

G=(contrary log (Y)-Wr*R-Wb*B)/Wg

Step 307: optionally be converted to other videos RGB and represent (alternatively, this step can also be performed as substituting of step 306) except appending to step 306.

Additional chrominance axis

In the image of dynamic range expanded and expanded contrast codomain, by means of the additional channel of visible and invisible two kinds of image informations, it may be useful increasing the visible wavelength passage.

The color gamut that can obtain from any given one group of redness, green and blueness does not comprise all possible visible color substantially.Primary colours such as synthetic red, green and blueness are features of human visual system to generate other ratios such as visible color such as yellow, orange, blue-green and brown, are called as " metamerism (metamerism) ".

Point out as the paper of being called of mentioning in the above " an expression example that is used for scalable, interoperable and extendible color of image and dynamic range ", might increase extra primary colours for red, green and blue three kinds of primary colours.Especially, blue-green (cyan), fuchsin (magenta) and yellow primary colours help to make the color gamut expansion, exceed by outside the most common obtainable scope of primary color values such as red, green and blue.In addition, can also transmit purple light and ultraviolet light (its make phosphorescent color shinny).

The verified invisible infrared wavelength beyond visible color is for passing cloud and mist and haze and seeing that in the dark thing is effective.Ultraviolet wavelength can be effective to the visual picture details of seeing low amplitude such as fingerprint and face coat also.

In addition, even in visible wavelength, different material (as smog and marine alga under water) also can reduce the dynamic range of numerical value or some wavelength of contrast usually.Here it is, and why smog can manifest brown, provides shade of brown for a long way off all objects, has reduced blue contrast and dynamic range.This also is that why underwater photography can shows green, dark green or blue, because the red end of visible spectrum is reducing aspect contrast and the dynamic range.

Aforesaid logarithmic relationship between Y, U and V will make the coding optimization for the color relationship of visible light.

In this aspect of the invention, extra chrominance channel is increased the passage to three fundamental wavelengths of coding, and this typically represents to realize with RGB or YUV.In addition, when using yuv color space, the formation that also may change Y (brightness) passage is beneficial to the picture signal of high amplitude.Therefore, for instance, along with luminance shifting arrives other wave-length coverages, green viewability channel can use the chrominance channel of himself to encode.Red, green and blue (perhaps other visible and invisible primary colours) each all have under the situation of chrominance channel of himself, this notion can be expanded to Y brightness be infrared ray.

According to this aspect of the invention, for each new chrominance channel, should determine following content:

Should 1) passage be encoded discriminatively with one or more other passages (normally with such as U=R-Y brightness)?

2) whether passage should be given full resolution with respect to brightness, can perhaps the intended purpose to a supposition reduce resolution and not damage picture quality?

1) in the correlation of determining to be based on each coding pass and other passages.For example, ultraviolet ray or far infrared wavelength image may be irrelevant relatively for visible wavelength or for each other.In this case, these passages can need not to relate to other passages and be encoded.But all visible wavelengths are height correlations, and it is almost always useful therefore relative to each other to encode usually.

Determine based on these, can select one group and exceed usually the image channel of (or substitute and exceed) three basic passages (for example YUV).For example, selected this group image passage can comprise a Y ' luminance channel, and n chrominance channel, as U ' first chrominance channel, V ' second chrominance channel and X ' the 3rd chrominance channel.

Utilize this example and use motion compensation compression, the Y ' value of selection will be encoded with full resolution, and various other chrominance channels (U ', V ', X ') will be encoded distinctively or independently.Unless have additional passage, all passages can utilize identical motion vector and macroblock motion compensation structure, this is used to conventional YUV with them and represents identical.Each this passage will with respect to Y utilize appropriate resolution (its with step 2 in the above determine identical).In addition, a QP biasing (as mentioned above) can be put on each chrominance channel independently, to guarantee to realize desired compressed chromaticity quality.

Even when only putting on visible wavelength, additional chrominance channel also can not only be guaranteed the color gamut and the more accurate color that are expanded, but also additional transparency, definition and noise fidelity are applied in to the color such as fuchsin, orange, yellow and this class highly-visible of water cyan (aquacyan).These advantages are particular importances for wide dynamic range and wide contrast range images.

Fig. 4 is one a flow chart, and it shows an illustrative method (its can by computer realization) of the additional chrominance channel that is used to encode in image compression system:

Step 401: in an image compression system, determine that an optimal brightness that is used for image represents that this image is selected based on wide dynamic range and the highest resolution, comprise selectable invisible light wavelength graph image signal.

Step 402: determine n additional chrominance channel with display image, wherein n is at least 3.

Step 403: optionally,, determine whether to be useful with respect to brightness and/or one or more other chrominance channel codings discriminatively for each chrominance channel.

Step 404: to each chrominance channel picture signal, determine desired resolution, make resolution be equal to or less than the resolution of brightness, and can select to implement resolution and reduce with respect to the luminance picture signal from input.

Step 405: utilize motion compensation to compress this Y+n chromatic diagram image signal.

Step 406: this Y+n chromatic diagram picture (normally in different computers) decompresses.

Step 407: reduce if implement resolution, then recover the original resolution of chrominance channel.

Step 408: with each chrominance channel and its different corresponding chrominance channel combination (if any) in step 403 in the above.

Step 409: optionally carry out following arbitrary step:

A) chrominance channel is transformed into a real space, perhaps is transformed into space, and regard a true color image as with three above primary colours such as RGB;

B) execution conversion a), but regard a pseudo color image (as infrared light is mapped as green light) as;

C) use chrominance channel and for handling and/or analyzing and do conversion.

As another selection, each chrominance channel can be with respect to the QP value used to luminance channel, QP value with biasing of using (increase or reduce), thus each chrominance channel is obtained the credit rating (promptly the trade off chrominance noise and the higher degree of compression) of expectation.

Embodiment

The present invention can realize with hardware or software or both combinations (for example programmable logic array).Unless otherwise specified, the algorithm that comprises as a part of the present invention is not relevant with any specific computer or other devices inherently.Especially various common apparatus can perhaps be can constitute isolated plant (as integrated circuit) to carry out specific function more easily by means of written program is used according to instruct herein.Therefore, the present invention can realize that this programmable computer system comprises at least one processor, at least one data-storage system (comprising easily become estranged nonvolatile memory and/or memory cell), at least one input unit or port separately, reaches at least one output device or port with one or more computer programs that are executed on one or more programmable computer system.Program code is used to import data, to carry out the function of describing herein and to produce output information.This output information is used for one or more output equipments in known manner.

Each said procedure can be communicated by letter with a computer system with any required computer language (comprising machine language, assembler language or high level procedural, logical language or object oriented programming languages).In a word, described language can be a kind of language that is compiled or explained.

Each aforementioned calculation machine program preferably be stored or download to one by the readable storage medium of general or special purpose programmable calculator or device (for example solid-state memory or for magnetic or be the medium of optical medium) on, with this storage medium of box lunch or device by this computer system reads time configuration and operational computations machine, thereby carry out the step of description herein.System of the present invention also can be conceived to realize by means of the computer-readable recording medium of computer program configuration as one, wherein storage medium is configured to such an extent that make this computer system with specific working with predetermined mode, to carry out the function of describing herein.

A plurality of embodiment of the present invention has been described.Yet should be understood that not breaking away from the spirit and scope of the present invention can carry out various modifications.For example, some step described above can sort separately, therefore can carry out to be different from described order.Therefore, other embodiment fall within the scope of the appended claims.

Claims

1. a method that is used in the image chroma information of video image compression system compressed color video image comprises that selection is used for the resolution of the red composition of this color video frequency image, makes it be higher than the resolution of the blue composition that is used for this color video frequency image.

2. method that is used in the image chroma information of video image compression system compressed color video image comprises:

(a), the descending filtration of blue composition of this color video frequency image is become the blue composition of oneself processing with first resolution along the level of this color video frequency image and at least one in the vertical image axle; With

(b) the red composition of this color video frequency image is filtered become the red composition of having handled with second resolution that is higher than this first resolution.

3. according to the method for claim 2, wherein along the level of this color video frequency image and at least one in the vertical image axle, this second resolution is in 0.5 to 1.0 the scope of full resolution of this redness composition.

4. according to the method for claim 2, further comprise: be somebody's turn to do oneself blueness and the red composition of processing to the major general and be collapsed into the compression output image.

5. according to the method for claim 4, further comprise: should compress output image and decompress to obtain unpressed processing back blueness and red composition.

6. according to the method for claim 5, further comprise: will handle back blueness and red composition and enlarge the full resolution that is filled into this color video frequency image.

7. according to the method for claim 1 or 2, wherein this video image compression system is a kind of motion compensated video image compression system.

8. method that is used for utilizing quantization parameter (QP) between compression period between the compression period of the color video frequency image of yuv video image compression system, reducing chrominance noise, this method comprises: the Y color channel for color video frequency image utilizes a QP value, and in the U of this color video frequency image and the V color passage one of them utilizes the 2nd QP value at least, wherein the 2nd QP value is less than a QP value.

9. method according to Claim 8, wherein the 2nd QP value is determined by a QP value is applied a bias.

10. method according to Claim 8 further comprises: after applying this first and second QP value, this color video frequency image is collapsed into the compression output image.

11. the method according to claim 10 further comprises: use this this compression output image of first and second QP value decompression, to obtain unpressed video image.

12. one kind is used for utilizing quantization parameter (QP) and realizes the more method of high compression between the compression period of the color video frequency image of yuv video image compression system, this method comprises: the Y color channel for color video frequency image utilizes a QP value, and utilize the 2nd QP value for the U of this color video frequency image and in the V color passage at least one, wherein the 2nd QP value is less than a QP value.

13. according to the method for claim 12, wherein the 2nd QP value is determined by a QP value is applied a bias.

14. the method according to claim 12 further comprises: after applying this first and second QP value, this color video frequency image is collapsed into the compression output image.

15. the method according to claim 14 further comprises: use this this compression output image of first and second QP value decompression, to obtain unpressed video image.

16. the method for code efficiency that is used for improving in video image compression system the color space of the video image be expressed as linear rgb pixel value at first comprises that the logarithm that linear rgb pixel value with this video image is converted into brightness and chrominance channel information represents.

17. according to the method for claim 16, conversion wherein comprises that using following formula represents with the YUV logarithm that obtains this video image:

Ylog＝Log(Wr*R+Wg*G+Wb*B)

U chrominance channel=Log (R)-Ylog

V chrominance channel=Log (B)-Ylog

Wherein Wr, Wg and Wb are the linear weighted functions for redness, green and the blue composition of the brightness of this video image.

18. the method according to claim 17 further comprises: reduce the U that this YUV logarithm represents and the resolution of V chrominance channel.

19. the method according to claim 17 further comprises: the YUV logarithm of this video image is represented to be collapsed into compressed video image.

20. the method according to claim 19 further comprises: this compressed video image de-compressed into to the YUV logarithm that is resumed of this video image represent.

21. the method according to claim 20 further comprises: the YUV logarithm that is resumed of this video image is represented to be converted into linear rgb pixel value.

22. according to the method for claim 21, conversion wherein comprises uses following formula to obtain this linearity rgb pixel value:

The contrary log (Y+U) of R=

The contrary log (Y+V) of B=

G=(contrary log (Y)-Wr*R-Wb*B)/Wg.

23. a method that is used to improve the characteristics of image of color video frequency image in image compression system comprises:

(a) select one group of image passage to represent this color video frequency image, comprise a luminance channel and n chrominance channel, wherein n is 3 at least; With

(b) this luminance channel and this n additional chrominance channel are collapsed into compressed video image.

24. according to the method for claim 23, wherein at least one chrominance channel is represented the invisible light wavelength.

25. according to the method for claim 23, wherein this luminance channel is to have the highest dynamic range and the image channel of resolution.

26. the method according to claim 23 further comprises: be independent of each other passage and each chrominance channel of encoding.

27. the method according to claim 23 further comprises: with respect to another selected passage and each chrominance channel of encoding discriminatively.

28. the method according to claim 23 further comprises: the resolution that reduces at least one chrominance channel.

29. the method according to claim 23 further comprises: apply a quantization parameter (QP) value and give at least one chrominance channel, this at least one chrominance channel is to setover with respect to a QP value that imposes on this luminance channel.

30. computer program that is stored on the computer-readable medium, be used for image chroma information at video image compression system compressed color video image, this computer program comprises instruction, with so that the resolution of the red composition of this color video frequency image is selected in the computer permission, make it be higher than the resolution of the blue composition that is used for this color video frequency image.

31. a computer program that is stored on the computer-readable medium is used for the image chroma information at video image compression system compressed color video image, this computer program comprises instruction, uses so that computer:

32. according to the computer program of claim 31, wherein along the level of this color video frequency image and at least one in the vertical image axle, this second resolution is in 0.5 to 1.0 the scope of full resolution of this redness composition.

33. the computer program according to claim 31 further comprises: with this blueness of having handled and red composition are collapsed into the instruction of compression output image so that this computer is to the major general.

34. the computer program according to claim 33 further comprises: use so that this computer will compress the output image decompression to obtain the unpressed instruction of handling back blueness and red composition.

35. the computer program according to claim 34 further comprises: with so that this computer will be handled the instruction that afterwards blueness and the expansion of red composition are filled into the full resolution of this color video frequency image.

36. according to the computer program of claim 30 or 31, wherein this video image compression system is a kind of motion compensated video image compression system.

37. computer program that is stored on the computer-readable medium, be used for utilizing quantization parameter (QP) between compression period between the compression period of the color video frequency image of yuv video image compression system, reducing chrominance noise, this computer program comprises in order to make computer utilize the instruction of a QP value and the 2nd QP value, the one QP value is used for the Y chrominance channel of color video frequency image, and the 2nd QP value be used for the U of this color video frequency image and V chrominance channel at least one of them, wherein the 2nd QP value is less than a QP value.

38. according to the computer program of claim 37, wherein the 2nd QP value is determined by a QP value is applied a bias.

39. the computer program according to claim 37 further comprises:

With so that this computer after applying the first and second QP values, with this color video frequency image be collapsed into the compression output image instruction.

40. the computer program according to claim 39 further comprises: with so that the decompression calculations machine uses this first and second QP value this compression output image that decompresses, to obtain the instruction of unpressed video image.

41. computer program that is stored on the computer-readable medium, be used for utilizing quantization parameter (QP) between compression period between the compression period of the color video frequency image of yuv video image compression system, obtaining higher compression, this computer program comprises and is used to make computer to utilize the instruction of a QP value and the 2nd QP value, the one QP value is used for the Y chrominance channel of color video frequency image, and the 2nd QP value be used for the U of this color video frequency image and V chrominance channel at least one of them, wherein the 2nd QP value is greater than a QP value.

42. according to the computer program of claim 41, wherein the 2nd QP value is determined by a QP value is applied a bias.

43. the computer program according to claim 41 further comprises:

With so that this computer after applying this first and second QP value, with this color video frequency image be collapsed into the compression output image instruction.

44. the computer program according to claim 41 further comprises: with so that the decompression calculations machine uses this first and second QP value this compression output image that decompresses, to obtain the instruction of unpressed video image.

45. computer program that is stored on the computer-readable medium, be used for improving the code efficiency that the color space of the video image be expressed as linear rgb pixel value is at first represented in video image compression system, this computer program comprises in order to make computer that the linear rgb pixel value of this video image is converted into the instruction that the logarithm of brightness and chrominance channel information is represented.

46. according to the computer program of claim 45, wherein usefulness is so that the instruction of this computer conversion comprises that usefulness is so that the instruction that the following formula of computer application is represented with the YUV logarithm that obtains this video image:

Ylog＝Log(Wr*R+Wg*G+Wb*B)

U chrominance channel=Log (R)-Ylog

V chrominance channel=Log (B)-Ylog wherein Wr, Wg and Wb is a linear weighted function for redness, green and the blue composition of the brightness of this video image.

47. the computer program according to claim 46 further comprises: with so that this computer reduces the instruction of the resolution of U that this YUV logarithm represents and V chrominance channel.

48. the computer program according to claim 46 further comprises: use so that this computer is represented the YUV logarithm of this video image to be collapsed into the instruction of compressed video image.

49. the computer program according to claim 48 further comprises: use so that the video image that the decompression calculations machine will compress de-compresses into the instruction of representing for the YUV logarithm that is resumed of this video image.

50. the computer program according to claim 49 further comprises: use so that this decompression calculations machine is represented the YUV logarithm that is resumed of this video image to be converted into the instruction of linear rgb pixel value.

51. according to the computer program of claim 50, wherein with so that the instruction of this computer conversion comprise with so that the following formula of computer application to obtain the instruction of this linearity rgb pixel value:

The contrary log (Y+U) of R=

The contrary log (Y+V) of B=

G=(contrary log (Y)-Wr*R-Wb*B)/Wg.

52. a computer program that is stored on the computer-readable medium is used for improving at image compression system the characteristics of image of color video frequency image, this computer program comprises instruction, with so that computer:

53. according to the computer program of claim 52, wherein at least one chrominance channel is represented the invisible light wavelength.

54. according to the computer program of claim 52, wherein this luminance channel is to have the highest dynamic range and the image channel of resolution.

55. the computer program according to claim 52 further comprises: use so that computer is independent of each other passage and the instruction of each chrominance channel of encoding.

56. the computer program according to claim 52 further comprises: with so that computer with respect to another selected passage and the instruction of each chrominance channel of encoding discriminatively.

57. the computer program according to claim 52 further comprises:

With so that computer reduces the instruction of the resolution of at least one chrominance channel.

58. the computer program according to claim 52 further comprises:

With so that computer applies the instruction that a quantization parameter (QP) value is given at least one chrominance channel, this at least one chrominance channel is to setover with respect to a QP value that imposes on this luminance channel.

59. a system that is used in the image chroma information of video image compression system compressed color video image comprises:

(a) be used to select this color video frequency image red composition resolution and make it be higher than the device of the resolution of the blue composition that is used for color video frequency image; With

(b) be used to apply selected resolution to compress the device of this color video frequency image.

60. a system that is used in the image chroma information of video image compression system compressed color video image comprises that device is used for:

(a), the descending filtration of blue composition of this color video frequency image is become the blue composition of having handled with first resolution along the level of this color video frequency image and at least one in the vertical image axle; With

61. according to the system of claim 60, wherein along the level of this color video frequency image and at least one in the vertical image axle, this second resolution is in 0.5 to 1.0 the scope of full resolution of this redness composition.

62. the system according to claim 60 further comprises: be used for the device that to the major general this blueness of having handled and red composition are collapsed into the compression output image.

63. the system according to claim 62 further comprises: be used for this compression output image is decompressed to obtain the unpressed device of handling back blueness and red composition.

64. the system according to claim 63 further comprises: the device that is used for this processing back blueness and red composition are enlarged the full resolution that is filled into this color video frequency image.

65. according to the system of claim 59 or 60, wherein this video image compression system is a kind of motion compensated video image compression system.

66. a system that is used for utilizing quantization parameter (QP) between compression period reducing chrominance noise between the compression period of the color video frequency image of yuv video image compression system comprises:

(a) be used to utilize the device of a QP value and the 2nd QP value, the one QP value is used for the Y chrominance channel of color video frequency image, and the 2nd QP value be used for the U of this color video frequency image and V chrominance channel at least one of them, wherein the 2nd QP value is less than a QP value; With

(b) be used between this color video frequency image compression period, applying the device of the QP value of this selection.

67. according to the system of claim 66, wherein the 2nd QP value is determined by a QP value is applied a bias.

68. the system according to claim 66 further comprises: be used for after applying this first and second QP value, this color video frequency image be collapsed into the device of compression output image.

69. the system according to claim 68 further comprises: be used to use this this compression output image of first and second QP value decompression, to obtain the device of unpressed video image.

70. one kind is used to realize the more system of high compression utilizing quantization parameter (QP) between compression period during the compressed color video image in the yuv video image compression system, comprising:

(a) be used to utilize the device of a QP value and the 2nd QP value, the one QP value is used for the Y chrominance channel of color video frequency image, and the 2nd QP value be used for the U of this color video frequency image and V chrominance channel at least one of them, wherein the 2nd QP value is greater than a QP value;

(b) be used between the compression period of this color video frequency image, applying the device of selected QP value.

71. according to the system of claim 70, wherein the 2nd QP value is determined by a QP value is applied a bias.

72. the system according to claim 70 further comprises: be used for after applying this first and second QP value, this color video frequency image be collapsed into the device of compression output image.

73. the system according to claim 72 further comprises: be used to use this this compression output image of first and second QP value decompression, to obtain the device of unpressed video image.

74. a system that is used for improving in video image compression system the code efficiency that the color space of the video image be expressed as linear rgb pixel value at first represents comprises:

(a) be used for the device of the linear rgb pixel value of inputted video image; With

(b) the linear rgb pixel value that is used to change this video image is the device that the logarithm of brightness and chrominance channel information is represented.

75. according to the system of claim 74, conversion wherein comprises that using following formula represents with the YUV logarithm that obtains this video image:

Ylog＝Log(Wr*R+Wg*G+Wb*B)

U chrominance channel=Log (R)-Ylog

V chrominance channel=Log (B)-Ylog

76. the system according to claim 75 further comprises: the device that is used to reduce the resolution of U that this YUV logarithm represents and V chrominance channel.

77. the system according to claim 75 further comprises: the device that is used for the YUV logarithm of this video image is represented to be collapsed into compressed video image.

78. the system according to claim 77 further comprises: be used for this compressed video image is de-compressed into the device of representing for the YUV logarithm that is resumed of this video image.

79. the system according to claim 78 further comprises: the device that is used for the YUV logarithm that is resumed of this video image is represented to be converted into linear rgb pixel value.

80. according to the system of claim 79, conversion wherein comprises uses following formula to obtain this linearity rgb pixel value:

The contrary log (Y+U) of R=

The contrary log (Y+V) of B=

G=(contrary log (Y)-Wr*R-Wb*B)/Wg.

81. a system that is used for improving at image compression system the characteristics of image of color video frequency image comprises device:

82. 1 system according to Claim 8, wherein at least one chrominance channel is represented the invisible light wavelength.

83. 1 system according to Claim 8, wherein this luminance channel is to have the highest dynamic range and the image channel of resolution.

84. 1 system further comprises: be used to be independent of each other passage and the device of each chrominance channel of encoding according to Claim 8.

85. 1 system according to Claim 8 further comprises: be used for respect to another selected passage and the device of each chrominance channel of encoding discriminatively.

86. 1 system according to Claim 8 further comprises: the device that is used to reduce the resolution of at least one chrominance channel.

87. 1 system according to Claim 8 further comprises: be used to apply the device that a quantization parameter (QP) value is given at least one chrominance channel, this at least one chrominance channel is to setover with respect to a QP value that imposes on this luminance channel.